Robot Writing Model With Human Writing Preference Based On Gated Recursive Unit

The research of robot writing is not only conducive to carry forward the traditional culture of Chinese calligraphy,but also can provide guidance for robots to achieve other creative and complex tasks independently.At present,there are still some problems in the research methods of robot writing,such as the lack of diversity of writing results and the writing style does not conform to human writing habits.In this paper,in order to improve the diversity of robot writing results and improve its writing style,we use the variational self encoder(VAE)and gating recursive unit(GRU)technology to create a method that enables the robot to write high-quality and diverse strokes in line with human writing habits.Firstly,a robot writing system based on variational self encoder is constructed.Because the writing process of robot is difficult to be represented by mathematical expression,the traditional gradient back propagation method can not be directly used to optimize the system network.Therefore,this paper also discusses the possibility of using competitive group optimization(CSO)algorithm to optimize the robot writing system.The system can write nine different strokes of Chinese characters,and the writing results have a certain diversity.This paper also proposes a robot writing system based on the combination of VAE and GRU.In addition,this study also proposes a new evaluation index of robot writing,which is incorporated into the writing feedback system.The index aims to help the robot learn to write in a variety of ways.The experimental results show that the robot writing system proposed in this paper can write numbers in line with human writing habits.The fr é Chet perception distance(FID)method is used to prove that the new evaluation index can bring diversity to the robot writing results,and the correlation between the robot writing results and the training samples is also proved.The system has the potential to develop autonomous behavior in other robot applications.